Multiple-camera System Research Articles

A Korean Cattle Weight Prediction Approach Using 3D Segmentation-Based Feature Extraction and Regression Machine Learning from Incomplete 3D Shapes Acquired from Real Farm Environments

Accurate weight measurement is critical for monitoring the growth and well-being of cattle. However, the traditional weighing process, which involves physically placing cattle on scales, is labor-intensive and stressful for the animals. Therefore, the development of automated cattle weight prediction techniques assumes critical significance. This study proposes a weight prediction approach for Korean cattle using 3D segmentation-based feature extraction and regression machine learning techniques from incomplete 3D shapes acquired from real farm environments. Firstly, we generated mesh data of 3D Korean cattle shapes using a multiple-camera system. Subsequently, deep learning-based 3D segmentation with the PointNet network model was employed to segment 3D mesh data into two dominant parts: torso and center body. From these segmented parts, the body length, chest girth, and chest width of Korean cattle were extracted. Finally, we implemented five regression machine learning models (CatBoost regression, LightGBM, polynomial regression, random forest regression, and XGBoost regression) for weight prediction. To validate our approach, we captured 270 Korean cattle in various poses, totaling 1190 poses of 270 cattle. The best result was achieved with mean absolute error (MAE) of 25.2 kg and mean absolute percent error (MAPE) of 5.85% using the random forest regression model.

Remote wide angle view broad wavelength viewing system compatible with D-T operations in JET

Imaging diagnostics in Joint European Torus have become essential in the study of plasma wall interactions and the protection of the plasma facing components. During deuterium operations, the location of these diagnostics was at close proximity to the vessel. However, for the 50–50 deuterium-tritium operation, the neutron yield impact onto the electronics of these imaging diagnostics would have caused them to fail completely at these locations. Shielding these systems from neutrons is achieved by relocating a selected number of these imaging systems to outside the Torus Hall (i.e. to the other side of the biological shield wall). The relocation encompasses the construction of two new lines of sight crossing through the biological shield wall and each of them using an optical relay which extends to separate ports in the machine. This article starts with a brief description of two new viewing systems. However, it focuses on the optical modelling and engineering design of one of them which is a multiple camera system producing a wide angle view (WAV) of the inside of the vessel. This WAV system has been used successfully during the recent D-T campaign. It operates in a broadband wavelength range extending from the visible to the middle infrared. It also incorporates a novel in-house horizontal kinematic mirror mount made from glass reinforced plastic as part of the optical relay. The operational stability of the WAV system has also been assessed and some post design improvements implemented.

Wide-viewing holographic stereogram based on self-interference incoherent digital holography.

We propose a holographic stereogram synthesis method which uses holograms that are optically captured by self-interference incoherent digital holography (SIDH) based on a geometric phase lens. SIDH is a promising solution for hologram acquisition under low-coherence lighting condition. A mechanical scanning system is constructed to acquire multiple perspective holograms. Numerical simulations and experimental analyses conducted using high-resolution diffractive optical element demonstrate that the proposed method can produce a wide-viewing hologram which can realize realistic 3D scenarios with depth cues such as accommodation and motion parallax. The future objectives include the implementation of a multiple-camera system for holographic videos.

A Two-Step Optimization for Extrinsic Calibration of Multiple Camera System (MCS) Using Depth-Weighted Normalized Points

A multiple camera system (MCS), which allows only a limited common field-of-view between adjacent cameras, has been chiefly calibrated using a typical 2D-3D point correspondence. However, the correspondence contains a potential instability by which the calibration result can diverge, and the instability has not been studied nor overcome before. We propose a MCS extrinsic calibration method with high robustness and accuracy based on a two-step optimization strategy. Using depth-weighted normalized points, we develop two novel types of point correspondence as follows. The 1st correspondence, F <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">cb</sub> , aims to robustly estimate the MCS extrinsic parameters by overcoming the potential instability that exists in the typical 2D-3D point correspondence. The 2nd correspondence, F <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">cc</sub> , aims to refine the MCS extrinsic parameters by using the direct relation between adjacent cameras. In the simulation, we validated the robustness and high accuracy of the proposed method. We validated its high precision in the field test. In both the simulation and the field test, our method was compared with the state-of-the-art method.

Semantically Synchronizing Multiple-Camera Systems with Human Pose Estimation.

Multiple-camera systems can expand coverage and mitigate occlusion problems. However, temporal synchronization remains a problem for budget cameras and capture devices. We propose an out-of-the-box framework to temporally synchronize multiple cameras using semantic human pose estimation from the videos. Human pose predictions are obtained with an out-of-the-shelf pose estimator for each camera. Our method firstly calibrates each pair of cameras by minimizing an energy function related to epipolar distances. We also propose a simple yet effective multiple-person association algorithm across cameras and a score-regularized energy function for improved performance. Secondly, we integrate the synchronized camera pairs into a graph and derive the optimal temporal displacement configuration for the multiple-camera system. We evaluate our method on four public benchmark datasets and demonstrate robust sub-frame synchronization accuracy on all of them.

Comparative Analysis of Soccer Performance Intensity of the Pre-Post-Lockdown COVID-19 in LaLiga™.

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) forced a stoppage in the 2019/2020 season of LaLiga™, possibly influencing performance indicators in the return to competition. Therefore, here, we evaluated whether the stoppage due to the coronavirus 2019 disease (COVID-19) lockdown influenced physical performance compared to the start of LaLigaTM in terms of high-intensity efforts. Using a semi-automatic, multiple-camera system, running activities during 22 matches were analyzed. We compared the first 11 matches of the season (pre-lockdown) with the 11 matches just after the restart of LaLiga™ (post-lockdown). The results showed higher (p < 0.05) performance in the pre-lockdown period compared with the post-lockdown period, including in medium-speed running (14.1–21 km/h), high-speed running (21.1–24 km/h), and sprinting speed running distances (>24 km/h). However, the number of accelerations/min and decelerations/min were significantly higher during the post-lockdown period. Therefore, we conclude that the stoppage due to the COVID-19 lockdown generated lower physical performance in the post-lockdown period compared with the pre-lockdown period, most likely due to the accumulation of matches (congested schedules).

航空多光谱相机三坐标无源定位技术

A large FOV infrared multiple spectrum camera, which contains advance infrared focal plane, acquires large FOV & high resolution multiple spectrum image toward area below the flight vehicle. It takes images on a whole day and nearly a whole whether condition. The multiple spectrum camera system configuration was proposed. Its angular measurement, three coordinate passive location key technology and solution were described in detail. The angular measure electro-circuit & clock synchronization should be finely designed. When the system had been integrated, angular error calibration and compensation, multi-sensor installation attitude calibration were proposed to improve angular measure accuracy and location accuracy. Flight tests shows that image mosaicking accuracy is 10″ which means the technic is effective.

A Multiple Video Camera System for 3D Tracking of Farmed Fry in an Aquaculture Tank

A Multiple Video Camera System for 3D Tracking of Farmed Fry in an Aquaculture Tank

High accuracy linear method for axis measurement

The attitude parameter is an important state parameter for a long axisymmetric target. The plane intersection (PI) method is a commonly used method for attitude estimation. However, this method only uses planes’ information in the object space under a multiple camera system (more than two cameras simultaneously observing a target). We propose two methods to address the aforementioned issue. One method involves minimizing the square of the object-space angle residual (OAR) and the other method involves minimizing the square of the image-space angle residual (IAR). The linear optimization method is used for the above minimizing problems. The simulation results demonstrate that the IAR method has higher accuracy than the PI and OAR methods under multiple and dual camera systems because it incorporates information of a pair of corresponding image points. Furthermore, our experiments have shown that the linear method is generally faster, and it has an equivalent accuracy compared to the iterative method.

Precise 3D Baseball Pitching Trajectory Estimation Using Multiple Unsynchronized Cameras

We developed a method for the precise estimation of the 3D trajectory of a baseball by modeling the movement of the baseball and estimating the capture delay, using multiple unsynchronized cameras. To develop the proposed algorithm, we mimicked the real-world process of capturing a baseball in simulation space, and analyzed the capture process using a multiple unsynchronized camera system. We represented the movement of the baseball using a piece-wise spline model, and predicted the position of the baseball in the subframes in a manner which is robust to position error and change in direction of movement of the baseball. This method accurately predicts the baseball position over time by modeling the movement of the baseball in a real baseball game environment, and improves the accuracy of the reconstructed 3D baseball trajectories. We defined an objective function to estimate the capture delay, and estimate the optimal capture delay parameter using non-linear optimization method. In addition, we evaluated the performance of the proposed method in simulation space and in a real-world situation. The experimental results show that the proposed method can estimate a 3D baseball trajectory precisely using a multiple unsynchronized camera system and is robust to variations in capture delay, both in the simulation space and in real-world situations.

Body scanning oriented 3D object modelling method based on multi-view depth camera

To implement the human body scanning process, in this paper, we propose a novel 3D object modelling method based on multi-view depth camera. The main innovations of this paper lie in that we model 3D object via fusing several depth images captured from depth cameras. Different from the 3D scanner-based 3D object modelling, this proposed method is much cheaper and easier to operate. To construct the multiple depth camera system, a camera is located at the centre point, and others are positioned at four directions with the same intervals. Afterwards, a cost function is defined to fuse the multiple views of depth images based on depth map of the surface points. Experimental results demonstrate that the proposed method can effectively 3D scan the human body, and our method is slightly inferior to 3D scanner. However, the proposed depth camera based method is much cheaper and easier to master.

Probabilistic Kinematic Model of a Robotic Catheter for 3D Position Control.

Continuum robots offer compliant and dexterous operations, which are suitable to be used in unstructured environments. Tendon-driven catheters, owing to their continuum structure, are applied in minimal invasive surgeries such as intracardiac interventions. However, due to the intrinsic nonlinearities and external disturbances, it is still a challenging task to accurately steer the catheter tip to the desired 3D positions. In this article, we proposed a new probabilistic kinematic model and a model-based three-dimensional position control scheme for a tendon-driven cardiac catheter. A dynamic Gaussian-based probabilistic model is developed to learn a mapping from the catheter states to the control actions. Based on the probabilistic model, a closed-loop position control is developed, in which the catheter is driven by a newly designed catheter driver system and tracked by a multiple near-infrared camera system. The proposed catheter framework is evaluated by the 3D trajectory tracking experiments both in a real 3D open space and in a minimum-energy-based simulator. The proposed control framework approximates the general kinematic by a combination of a catheter translation model and a distal workspace model, which provides the ability of automatically positioning the catheter tip in 3D and improving the accuracy by compensating the learned nonlinear effects.

Estimating Alpine Skiers' Energetics and Turn Radius Using Different Morphological Points.

Alpine ski analysis has always been very challenging, mainly due to the environmental conditions, large field and rapid and dynamic skiers’ movements. Global navigation satellite system (GNSS) offers a solution adapted to outdoor testing, but the relationship between the point where the antenna is attached and the real centre of mass (CoM) position is still unknown. This article proposes to compare different points of the body used to quantify the performance of alpine skiers. 3D models of seven elite skiers performing giant slalom (GS) were built using multiple camera system and dedicated motion tracking software. CoM as well as pelvis, head and feet trajectories were deduced from the data. The potential and kinetic energies corresponding to these points were calculated, as well as the evolution of the turn radius during the turn cycle. Differences between values given by the CoM and the other morphological points were analyzed. The pelvis offered the best estimation of the CoM: No differences were found for the biomechanical parameters, except for the kinetic energy, where 2% of the turn cycle had significant different values. The head was less accurate compared to the pelvis, showing significant differences with CoM between 7 and 20% of the turn cycle depending on the parameter. Finally, the feet offered the worst results, with significant differences between 16 and 41% of the turn cycle. Energies and turn radius calculated by using pelvis in place of CoM offered similar patterns, allowing the analysis of mechanical and dissipation energy in GS. This may potentially enable easier testing methods to be proposed and tested.

The Reliability of Technical and Tactical Tagging Analysis Conducted by a Semi-Automatic VTS in Soccer.

The Video Tracking multiple cameras system (VTS) is a technology that records two-dimensional position data (x and y) at high sampling rates (over 25 Hz). The VTS is of great interest because it can record external load variables as well as collect technical and tactical parameters. Performance analysis is mainly focused on physical demands, yet less attention has been afforded to technical and tactical factors. Digital.Stadium® VTS is a performance analysis device widely used at national and international levels (i.e. Italian Serie A, Euro 2016) and the reliability evaluation of its technical tagging analysis (e.g. shots, passes, assists, set pieces) could be paramount for its application at elite level competitions, as well as in research studies. Two professional soccer teams, with 30 male players (age 23 ± 5 years, body mass 78.3 ± 6.9 kg, body height 1.81 ± 0.06 m), were monitored in the 2016 season during a friendly match and data analysis was performed immediately after the game ended. This process was then replicated a week later (4 operators conducted the data analysis in each week). This study reports a near perfect relationship between Match and its Replication. R2 coefficients (relationships between Match and Replication) were highly significant for each of the technical variables considered (p < 0.001). In particular, a high score of interclass correlation and a small coefficient of variation were reported. This study reports meaningless differences between Match and its Replication (intra-day reliability). We concluded that the semi-automatic process behind the Digital.Stadium® VTS was more than capable of recording technical tagging data accurately.

A new image mosaicking approach for the multiple camera system of the optical remote sensing satellite GaoFen1

ABSTRACTThe GaoFen1 (GF1) optical remote sensing satellite is the first in China’s series of high-resolution civilian satellites and is equipped with four wide field of view (WFV) cameras. High accuracy image mosaicking depends on the geometrical quality of the stitched satellite imagery. This letter proposes a high accuracy image mosaicking approach based on combining multiple WFV cameras into a big virtual camera (BVC). In this method, a BVC is created according to the rigorous imaging model of a single camera, from which a final stitched image and corresponding high accuracy rational function model (RFM) can be obtained simultaneously. WFV images from GF1 and reference digital elevation models from ZiYuan3 are used to validate the proposed method. The experimental results validate that seamless mosaicking can be achieved directly from the geometrical relationship of the imaging model of the BVC and WFV camera. The proposed method is able to maintain the absolute and relative internal positioning accuracy of the original images.

Multiple View Depth Generation Based on 3D Scene Reconstruction Using Heterogeneous Cameras

In this paper, we introduce the multiple view depth generation method using heterogeneous cameras based on 3D reconstruction. The main goal of this research is to generate accurate depth images at each viewpoint of color cameras by using depth cameras placed at different positions. The conventional filter-based framework has critical problems such as truncated depth regions and mixed depth values. It degrades not only the quality of depth images but also synthesized intermediate views. A proposed framework is based on the 3D reconstruction method from the multiple depth cameras. The proposed system setup consists of two camera layers including four color cameras on a lower layer and two depth cameras on an upper layer as a parallel form. First, we estimate correct camera parameters using the camera calibration method on the offline process. In the online process, we capture synchronized color and depth images from the heterogeneous multiple camera system. Next, we generate 3D point clouds from 2D depth images and register them by the iterative closest points method. Then we can obtain an integrated 3D point cloud model. After that, we create the volumetric surface model from the sparse 3D point clouds by the truncated signed distance function. Finally, we can estimate the depth image at each color view by projecting the volumetric 3D model. In the experiment result and discussion section, we will verify not only the proposed framework resolves the aforementioned problems, but also has several advantages over the conventional framework

Walking Pattern Analysis Using an Acceleration Sensor Device

In this paper, a device to analyze gait pattern was developed by using a 2-axis acceleration sensor attached to the foot. The 1st low-pass filter was adapted to limit the frequency band up to 5 Hz. An algorithm to detect the peak value exceeding the threshold voltage of an X-axis acceleration sensor and a Z-axis acceleration sensor was developed and normal and abnormal walking patterns were thus differentiated. Also, MCU and Bluetooth were combined to transfer the data to other MCUs to display on an LCD; the size of the device could then be reduced. The new algorithm and the device allowed the individual walking patterns to be easily measured at a low cost and with less restriction on activities compared to conventional multiple pressure sensors or motion camera system.

Probabilistic Approach for Visual Homing of a Mobile Robot in the Presence of Dynamic Obstacles

An autonomous mobile robot returns back to a reference (home) position using just the visual information in the visual homing task. Challenges in this autonomous task involve efficient information extraction from images and executing motion based on this information. This paper aims to develop visual homing technique by extracting coarse location information with respect to the reference position using a novel bit encoding approach. Coarse information is used in probabilistic framework to build stronger beliefs on the robot's location with successive movement of the robot. The changes in visual information due to the presence of dynamic obstacles are accounted in likelihood information using bit encoding. Novel information encoding and proposed probability distribution develop sensor model. The action model aims for building stronger beliefs with successive movement, when limited number of commands are used. Experiments are performed using Fire Bird XI robot equipped with a multiple camera system for capturing panoramic view at the robot's location. Multiple experiments show that the robot reaches the reference position without the knowledge of initial location and changing environment conditions. This paper also reports experimental results in the presence of dynamic obstacles.

The effect of a succession of matches on the physical performance of elite football players during the World Cup Brazil 2014

The aim of the current study was to examine the influence of playing multiple games with a short recovery time between matches on physical activity in elite football. The variation of physical performance (overall distance, walking and light-intensity running, moderate-intensity running, high-intensity running and Sprint) of players during the group stage in the World Cup Brazil 2014 was examined. Footballers played three consecutive matches separated by 4 days. Altogether, 301 physical performance profiles were examined (Central Defenders (CD) = 57, External Defenders (ED) = 59, Central Midfielders (CM) = 61, External Midfielders (EM) = 56, and Forwards (F) = 68). Data were collected using a computerized multiple-camera tracking system (Prozone). No differences were found across the three matches in the distance covered in Sprint, HIR, MIR and Light-IR. External and Central Midfielders reported significant differences in the overall distance covered or that covered in MIR and HIR (p<0.05), however there is not a clear tendency of how playing multiple games modified their physical performance. In conclusion, fixture congestion had no, or very limited, influence on physical performance in elite football players. This indicates that, to a certain extent, top players can cope with a busy match schedule without underperforming.

Percentage of human-occupied areas for fall detection from two views

Falls are the major causes of fatal injury for the elderly population. To remedy this, several elderly people monitoring systems with fall detection functionality have been proposed. In this work, we investigate a video-based method of detecting fall incidents from multiple cameras. Our goal is to propose a novel method to detect falls on the floor with a multiple-camera system using the percentage of human-occupied areas. We suggest the use of two relatively orthogonal views to estimate the percentage of the surface of the person which is in contact with the ground according to the foreground information of each camera. These features are computed to differentiate by an automatic manner the lying on floor posture which can be considered as fall to other position such as standing up or sitting. This method is evaluated on a public multi-view fall detection dataset which contains videos of a healthy subject who performed 24 realistic scenarios. These scenarios show 22 fall events and 24 confounding events. The results of our experiments show that our proposed algorithm achieved 95.8 % sensitivity and 100 % specificity with less computational costs than state-of-the-art methods.